Self-reported hearing handicap in adults aged 55 to 81 years is modulated by hearing abilities, frailty, mental health, and willingness to use hearing aids

Abstract

Objective

The aim of this study was to predict outcomes of the HHI questionnaire (Hearing Handicap Inventory) using individual variables beyond pure-tone hearing thresholds.

Design

An extensive health-related test battery was applied including a general anamnesis, questionnaires, audiological measures, examination of visual acuity, balance, and cognition, as well as tactile- and motor skills. Based on the self-assessment of health variables and different sensory and cognitive performance measures, a frailty index was calculated to describe the health status of the participants. A stepwise linear regression analysis was conducted to predict HHI scores.

Study sample

A mixed sample (N = 212) of 55- to 81-year-old, participants with different hearing and aiding status completed the test battery.

Results

The regression analysis showed statistically significant contributions of pure-tone hearing thresholds, speech recognition in noise, age, frailty, mental health, and the willingness to use hearing aids on HHIE outcomes.

Conclusions

Self-reported hearing handicap assessed with the HHI questionnaire reflects various individual variables additionally to pure-tone hearing loss and speech recognition in noise. It is necessary to be aware of the influences of age and health-related variables on HHI scores when using it in research as well as in clinical settings.

Keywords:

• Aging
• behavioural measures
• HHIE
• hearing health

1. Introduction

Hearing impairment is one of the most prevalent disabilities, especially in old age. Usually, audiologists assess hearing ability through pure-tone hearing thresholds or percentage-word-correct speech-recognition tests. While these methods are essential for describing the functional capacity of the hearing organ and estimating any sensory restrictions, the results do not tell a complete story. Typical consequences of a hearing impairment include difficulties interpreting deteriorated speech signals, thus significantly affecting daily life or challenging independence, especially in old people (Dalton et al. 2003). Following the framework of the International Classification of Functioning, Disability and Health (ICF; WHO 2012), a long list of activity limitations and participation restrictions has been described that result from hearing impairment (Dixon et al. 2020). Several studies have demonstrated that hearing impairment is often associated with social isolation, reduced self-esteem, cognitive decline, depression, and anxiety (Lin et al. 2011; Li et al. 2014; Mick, Kawachi, and Lin 2014; Vos et al. 2016). However, the impact that hearing impairment has in the life of a person varies and, depending on its extent, it impairs emotional, physical, and social functions. Thus, hearing rehabilitation not only aims at restoring hearing ability, but through a patient-centered care at optimal outcomes regarding the person’s ability to participate in and to enjoy communication and activities of daily life, thus improving the general well-being of the person and their quality of life (QoL). With life expectancy increasing and more people living longer lives, understanding the impact of hearing impairment on QoL in old age is of great and increasing importance.

The assessment of QoL in the context of hearing rehabilitation is still a relatively new concept, and very few validated instruments are available. The Hearing Handicap Inventory for the Elderly (HHIE; Ventry and Weinstein 1982) is a well-established, hearing-related instrument that assesses QoL (Chisolm et al. 2007; Lazzarotto, Baumstarck, and Auquier 2016). Initially developed for elderly persons who are already retired, Newman et al. (1990) proposed a Hearing Handicap Inventory for Adults (HHIA) by substitution of three questions that are more appropriate for younger adults who are still in the work force. Both questionnaires (hereafter referred to as HHIE/A) include 25 items belonging either to an emotional or to a social/situational subscale. The HHIE/A is an important instrument to evaluate the impact of hearing loss (HL) and experience of handicap in a person’s everyday-life (Ciorba et al. 2012). In this manner, the HHIE/A has been regularly used to evaluate hearing aid (HA) outcomes in audiological intervention studies (see Aazh and Moore 2017, for an overview). The hearing handicap experienced was often reduced after aiding and adaptation to the amplification, indicating an improvement of hearing-related QoL (Malinoff and Weinstein 1989; Stark and Hickson 2004; Gopinath et al. 2012; Chang et al. 2016).

Research in hearing impairment and hearing-related QoL are especially relevant in old age. However, numerical age falls short of describing this heterogeneous group of persons, who are affected not only by sensory loss, but co-occurring health, social, and psychological changes, as well as a complex interaction of – often progressive or chronic – conditions over time. In gerontology and geriatrics, the concept of frailty has gained importance for describing this dynamic state, particularly as frailty has been shown to be linked to an increased risk for adverse outcomes such as poor QoL, and nursing home placement, as well as death within a year (Bilotta et al. 2011; Gobbens & van Assen, 2014). Frailty was defined as a “medical syndrome with multiple causes and contributors that is characterised by diminished strength, endurance, and reduced physiologic function that increases a person’s vulnerability for developing increased dependency and/or death” (Morley et al. 2013, p.2). Although a gold standard to measure frailty does not yet exist (Dent, Kowal, and Hoogendijk 2016), there are different approaches to measuring this concept. One approach is to use single measures that may indicate physical frailty (Kamil et al. 2016). Other studies examined the characteristics of a frailty phenotype, such as exhaustion, weight loss, low activity, a slow walk, and grip strength to characterise frailty status (Fried et al. 2001). A well-validated approach to measuring frailty by inclusion of various health-related variables (physical, cognitive, social) is the calculation of a frailty index of accumulated deficits (FI-CD, Dent, Kowal, and Hoogendijk 2016). For this purpose, 30 or more health-related variables are used to calculate the proportion of existing deficits and characteristics (Searle et al. 2008). While often not applicable in clinical settings (Dent, Kowal, and Hoogendijk 2016), the approach of cumulating deficits to calculate a frailty index was appropriate for the study reported, because it could be derived from existing data.

In many non-clinical audiological studies, a selection process takes place. Older persons with different degrees of HL voluntarily participate by responding to public announcements, e.g. in the local newspaper. After contact, e.g. by telephone, the potential candidates are usually invited to participate in the study at the respective institution. With this procedure, all participants meet the requirements set by the environmental conditions of the recruiting institution. Persons who are dependent on help due to physical or mental limitations are more likely to be unable to participate in non-clinical studies. The same applies to persons with reduced mobility. The groups of participants examined in hearing research therefore do not usually represent the general public. According to Ritchie et al. (2016), this bias in the selection of volunteers leads to certain characteristics of the participating groups. The younger participants in a gerontological sample may be motivated because they already experience age-typical problems in everyday life, while older participants in the sample may have fewer health deficits than their peers. This effect is not sufficiently captured by the numerical ages of the persons. Here again, the concept of frailty can be useful in assessing the aging state of a person irrespective of numerical age, while considering several domains of functioning.

The present study takes a broad approach in assessing factors that potentially influence QoL in the context of hearing impairment and hearing rehabilitation in older age. The dataset was established between November 2015 and July 2017 and comprises 223 participants covering 3 age decades that were recruited from a volunteer database and the general public (newspaper). In comparison to other research on the association of HL and QoL, for which clinical or convenience samples are often used, our sample also includes participants without associations to clinical settings or hearing-related care. The data set includes objective measures of sensory abilities (hearing, vision, dexterity), physical capabilities (psycho-motor, balance, health), and speech recognition and cognition (language, attention, memory), which together allow for a comprehensive evaluation of a person’s overall health and frailty. To complete the picture, subjective ratings were also collected, from the perspective of the participant, regarding their hearing handicap and emotional health (depression, anxiety), and reflecting self-perceived communication difficulties as well as social and emotional well-being.

Based on published findings highlighting the relationship between QoL and HL, this study aimed to determine individual factors that are related to self-reported hearing handicap and thus hearing-related QoL in old age as measured with the HHIE/A questionnaire. Furthermore, a FI-CD approach was used to assess frailty, targeted to describe aging beyond numerical age and its associations with the hearing handicap experienced.

2. Material and methods

2.1. Study design

Participants were recruited from the volunteers database of the Hörzentrum Oldenburg GmbH (n = 77) as well as via newspaper and public announcements (n = 146). A total of 223 participants completed 2 sessions on the premises of the Jade University of Applied Sciences (Oldenburg, Germany), with a duration of approximately 2.5 h each. An hourly rate of 12 Euro was paid to compensate their effort and all participants gave their informed consent prior to inclusion in the study. The study was approved by the ethics committee (“Kommission für Forschungsfolgenabschätzung und Ethik”) of Carl von Ossietzky University in Oldenburg, Germany (Drs. 22/2014).

2.2. Participants

A qualified clinical neuropsychologist assessed the neuropsychological results of all participants regarding any pathological alterations. This led to the exclusion of seven participants from the sample accompanied by the recommendation to consult a specialist due to specific neurological deficits. Furthermore, four participants were excluded due to missing values in the emotional section of the HHIE/A questionnaire. The remaining 212 participants had a mean age of 67.1 years (SD = 7.3), and these were evenly distributed over the entire age range of 55 to 81 years (Table 1). The majority were already retired. HL varied widely across participants. The pure-tone average HL in the audiometric test frequencies 0.5, 1, 2, and 4 kHz of the better ear was calculated and classified based on the definition of the World Health Organisation (PTA4; WHO 1998). Furthermore, to classify the audiometric results the pure-tone-audiometric part of the German guidelines for hearing aid (HA) provision (G-BA 2017) were applied. Hence, hearing aids were provided to participants that had pure-tone thresholds of 30 dB HL or higher for at least one test frequency between 0.5 and 4 kHz. For these participants, amplification was ensured during all measurements, including interviews and neuropsychological testing. They were free to choose between wearing their own devices (if available) or provided hearing devices (Phonak Bolero V90-P/SP with slim tubes and domes) fitted to the individual HL using the NAL-NL2 formula of the manufacturer. Of the 124 participants fulfilling the guidelines, 68 participants owned hearing aids. Among them were all 5 participants with severe hearing loss and 26 out of 30 participants with moderate hearing loss. In total, 19 participants were fitted with Phonak Bolero V90-SP and the remaining 105 participants were fitted with Phonak Bolero V90-P.

Table 1. Description of the participants, including socio-demographic and hearing characteristics.

Parameter	n (% of total)
Age in groups
55–60 years	50 (23.6)
61–65 years	43 (20.3)
66–70 years	43 (20.3)
71–75 years	42 (19.8)
76–81 years	34 (16.0)
Gender
Female	116 (54.7)
Male	96 (45.3)
Employment status
Full-time	20 (9.4)
Part-time	18 (8.5)
Marginal	6 (2.8)
Unemployed	11 (5.2)
Retired	155 (73.1)
Not available	2 (0.9)
WHO HL classification (better-ear thresholds)
No HL (PTA4 ≤ 25 dB HL)	134 (63.2)
Mild (26 dB HL ≤ PTA4 ≤ 40 dB HL)	43 (20.3)
Moderate (41 dB HL ≤ PTA4 ≤ 60 dB HL)	30 (14.2)
Severe (61 dB HL ≤ PTA4 ≤ 80 dB HL)	5 (2.4)
German HA guideline classification
No HA indication	88 (41.5)
HA indication	124 (58.5)

2.3. Hearing handicap inventory for elderly/adults (HHIE/A)

The HHIE/A questionnaires (Ventry and Weinstein 1982; Newman et al. 1990) were used to assess the consequences of hearing impairment on daily life. The HHIE/A consists of 25 items each, divided into 2 subscales: social/situational (12 items) and emotional (13 items). The question “Are you retired?” was included in the questionnaire and participants were asked to answer the specific items of the “E” or “A” version based on their response. Participants were requested to respond by choosing one of three options underlying a ranked scoring: “yes” (4 points), “sometimes” (2 points), and “no” (0 points). The maximum sum score for the whole scale was 100. Higher sum scores indicated greater self-perceived hearing handicap. In this study, the German version developed by Tesch-Römer (2001) was used. The questionnaire was sent to the participants prior to the first appointment by mail; participants were instructed via telephone to complete the questionnaire at home and return it at their first session in the lab. Experimenters checked for completion of the whole questionnaire and offered support if needed. Of the participants, 73.1% were already retired and completed the HHIE questionnaire, 26.9% completed the HHIA questionnaire. Missing values and questions that were rated as “not applicable” in the subscale “social” were set to 0, as suggested by Ventry and Weinstein (1982). This was the case for 10 participants, of whom 8 did not respond to item S11 (“Does a hearing problem cause you to attend religious services less often than you would like?”). Two participants did not respond to items related to watching TV (S15, S23). This procedure is not adequate for the “emotional” subscale because all of these items are applicable (Ventry and Weinstein 1982) and there must be other reasons that led to the omissions, which cannot be traced. As stated earlier, four participants were therefore excluded due to missing values in the “emotional” subscale.

2.4. Audiological assessment

Pure-tone audiometry was carried out in a soundproof booth using a Unity 2 audiometer (Siemens Audiologische Technik GmbH) and circumaural headphones (HDA 200, Sennheiser GmbH) for air-conduction thresholds at 10 test frequencies between 0.125 and 8 kHz. As stated above, the PTA4 was calculated by averaging the better-ear thresholds at 0.5, 1, 2 and 4 kHz.

Speech recognition was tested in a soundproof booth. The German Goettingen Sentence Test (GÖSA, Kollmeier and Wesselkamp 1997) was administered using the Oldenburg Measurement Application (OMA, HörTech gGmbH), unaided for all participants and irrespective of the individual HL. The speech material consisted of everyday-life sentences containing 3–8 words. Stimuli were presented together with speech-shaped noise of 65 dB SPL from the front (0° azimuth) via loudspeaker at 1.2 m distance to the participants. The examiner selected the correctly repeated words on a touchscreen. The signal-to-noise ratio (SNR) for a speech recognition score of 50% (SRT50) was measured in an adaptive procedure (Brand and Kollmeier 2002), starting at an SNR of 0 dB.

2.5. General health

The medical history of the participants was recorded using a questionnaire that had been developed in-house and was based on common tools for collecting health-related data. It included a self-assessment and an interview part, and covered questions about language acquisition, general health, chronic diseases, medication, balance and motor skills, dizziness, hearing problems and interventions, tinnitus, and noise exposure, as well as questions about socio-demographics. Exact wording of questions that were used to calculate the FI-CD are reported in the supplement.

Health related QoL was integrated in the general-health part of the questionnaire. Four questions were adapted from the German version of the SF36 health survey for health-related QoL (No. 1, 9d, 9e and 9f; Bullinger 1995) that are also part of the short form SF12 (Gandek et al. 1998). One question covered physical health (see Supplement), and three questions mental health. The mental health ratings (“How much of the time during the past four weeks have you felt […]”, “[…]full of energy?”, “[…]calm and peaceful?”, “[…] downhearted and blue?”) were accumulated and are defined as a marker for emotional well-being in the following analysis.

2.6. Frailty index

To quantify frailty, the method of accumulating deficits according to Searle et al. (2008) was applied retrospectively to the existing data. Calculation of the FI-CD typically involves approximately 30 to 70 characteristics that must meet certain conditions. All variables must be associated with a health issue, the prevalence must increase with age, the saturation of their age dependency must not occur too early, and they must be distributed over a range of systems in order to become a general measure (Searle et al. 2008). In a first approach, 28 variables that fulfilled the criteria defined by Searle et al. (2008) were selected, and the FI-CD₂₈ was calculated (Nüsse et al., 2018). Most of the variables included were derived from the questionnaires, reflecting specific health problem (e.g. hypertension, diabetes, heart attacks, tinnitus, motor impairments), self-assessed health status (adapted from SF12, see above), and hearing abilities (SSQ scores). Furthermore, a range of behavioural data was included, being PTA4, the timed “Up & Go” test, a composite score of cognitive tests (working memory, verbal memory, executive functioning, word fluency, interference, and divided attention), fine motor skills, and tactile resolution. All variables were recoded following the suggestions of Searle et al. (2008) with “0” indicating the absence and “1” the presence of a deficit. Depending on the specific variable, cut-off points and rank scores were defined (see supplemental material for details). The FI-CD₂₈ was calculated as the average of all scores resulting in a value between 0 and 1, with higher values indicating more frailty. In a second approach, to restrict the frailty variable to non-hearing health issues and to avoid doubling in the regression model, the hearing-related variables (tinnitus question, PTA4, and SSQ scores), as well as the mental-health ratings, were removed from the calculation, resulting in the FI-CD₂₃.

2.7. Aiding status

The HA status was collected in an interview questionnaire. Therefore, the examiner asked, among other questions, about ownership, duration of use, current use, testing, and willingness to use HA. Depending on the question, a positive answer was followed by more detailed questions (e.g. satisfaction). Based on this information, participants were grouped for further analysis into those who were willing to test HAs, or already use HAs (n = 75), and non-users without an indication or willingness to use HA (n = 135). This variable is named willingness to use HA (HA willingness/use) in the following.

2.8. Other measures

Additionally to the above tests, different measures were collected in the study, which are not reported in this paper but should be briefly mentioned. In the audiological domain, these were bone-conduction thresholds (at 0.5, 1, 2, 4 and 6 kHz), uncomfortable levels (at 0.5, 1, 2 and 4 kHz), and supra-threshold auditory processing (IPD-FR Test; Holube et al. 2020). Moreover, speech recognition was tested in other aided configurations involving numbers in quiet, GÖSA in quiet, speech-shaped noise, speech masker (IFFM, Holube et al. 2010) and in a virtual cafeteria environment (Grimm, Luberadzka, and Hohmann 2019). A mobile setup was used to assess visual acuity, contrast sensitivity, and colour vision (Vistec AG). The outcomes were not included in the frailty index, because they did not fulfil the requirements of Searle et al. (2008), i.e. visual accommodation saturates at the age of 55. Balance examinations were only carried out using the Vertiguard system (Zeisberg GmbH) if participants reported no falls, no increased fall risk, or dizziness, in their medical history.

2.9. Procedure

After a personal contact by phone or e-mail to arrange an appointment, information about the study and self-report questionnaires were sent to the participants prior to the first session. Overall, three different persons were trained to conduct the study. The experimenter was not changed within participants. At the first session, general anamnesis took place, followed by the audiological assessments and the vision testing. The second session started with the balance testing and included the test of tactile perception, the examination of fine motor skills and neuropsychological tests (memory, divided attention, executive function). Participants conducted tests in an equal fixed order. To avoid fatigue, the order was defined aiming to alternate test conditions (e.g. motor, auditory, visual) during the sessions. In addition, breaks were included whenever needed.

2.10. Data analysis

Study data were analysed regarding descriptive statistics, group differences, and stepwise forward linear regression analysis. Descriptive characteristics and age dependency of all variables used in the regression model are shown in Table 2. Following literature findings (Weinstein and Ventry 1983; Newman et al. 1990), pure-tone hearing loss and recognition were included first to the model. Based on the hypothesised influence of age, general health and HA status on HHI scores, more predictors were included in the model. If available, different variables representing the hypothesised theoretical construct (e.g. aided and unaided SRTs for the construct speech recognition) were added parallelised to the model in order to identify the variable with highest relevance for HHI scores. IBM SPSS Statistics (Version 26.0.0.0) was used for all analyses.

Table 2. Proportions of gender distribution and hearing aid status per 5-year age cohort, with means and standard deviations (STD).

Age group	55–60 years	61–65 years	66–70 years	71–75 years	76–81 years
Males/females	18/32	21/22	21/22	21/21	15/19
HHIE/A
Mean	5.6	7.7	5.5	6.7	5.2
STD	12.8	9.6	6.6	7.1	5.6
PTA4 (dB HL)
Mean	11.9	23.8	23.4	31.0	30.7
STD	7.7	19.3	12.5	16.4	16.5
SRT50 (dB SNR)
Mean	−4.3	−3.1	−3.0	−2.1	−2.1
STD	0.9	1.4	1.6	2.2	1.9
FI-CD23
Mean	0.25	0.24	0.22	0.27	0.22
STD	0.10	0.10	0.11	0.11	0.09
Mental health
Mean	8.0	7.4	7.6	6.8	7.4
STD	2.1	2.3	2.1	1.4	2.6
HA willingness/use
n	2	18	15	23	17
(% of total)	4.0	41.9	34.9	54.8	50.0

3. Results

3.1. Frailty index

As a reduction of included variables in the calculation of the frailty index was necessary to prevent from covariations in the further analyses, the two approaches FI-CD₂₈ and FI-CD₂₃ were correlated to demonstrate their comparability. The initially calculated FI-CD₂₈ (Nüsse et al, 2018) showed a large variance in all age groups (Figure 1). No clear relationship between FI-CD₂₈/FI-CD₂₃ and the numerical age was observed in our data. After exclusion of variables (see Section 2.5), FI-CD₂₃ scores decreased on average by 0.026 points. Nevertheless, index scores of FI-CD₂₈ and FI-CD₂₃ correlated highly (r = 0.97, p < 0.001). Although FI-CD₂₃ contains only 23 variables, the high correlation implies that the index represents the cumulative deficits of the participants and will be used in the following analysis.

Figure 1. FI-CD in relation to numerical age: diamonds indicate the first FI-CD approach, with 28 variables; circles indicate the second FI-CD version, with 23 variables omitting hearing-related variables and mental-health scores. Data points of the same participants are connected by grey lines, means for each age are indicated by filled symbols and dashed lines.

3.2. Age-dependent descriptive analysis

The hearing-related characteristics of the 212 participants are presented in Table 2, divided into age groups of 5 (or 6) years. Both age and gender distribution are consistent in the data sample. PTA4 varied between −1 and 76 dB HL and slopes mainly showed high frequency HLs, as expected for elderly participants with presbyacusis. Higher values in HHIE/A scores indicate a greater self-perceived hearing handicap. Of the whole sample, 14.6% (n = 31) had a profound handicap indicated by an HHIE/A score of 26 and higher. Similarly, higher PTA4, SRT50 and FI-CD₂₃ indicate worse performance or higher frailty. Higher mental-health scores indicate better self-perceived mental health, following the scoring of the SF12 questionnaire. Overall, pure-tone thresholds, speech recognition scores, and the willingness to use HAs tended to increase with age. No age-related trend was found for HHIE/A scores, frailty index, or mental-health scores. In a pairwise comparison, no significant difference was found between the HHIE scores for participants who were retired and the HHIA scores for the actively working participants (Wilcoxon-Test, T = 4210.5, p = 0.496). Therefore, HHIE/A total scores are used in further analysis.

3.3. Regression analysis

A stepwise forward linear-regression analysis was conducted to identify influencing factors that model variance in the HHIE/A scores. Results of this analysis are reported in Table 3, which shows the model with the highest R². Overall, six independent predictors contributed significantly to the HHIE/A ratings, explaining 44% of the variance. Literature findings indicated a substantial relationship between pure-tone hearing loss and HHIE/A scores (Weinstein and Ventry 1983; Newman et al. 1990), thus this variable was included first, using the PTA4 of the better ear as predictor (R² = 0.29). In a second step, speech recognition was included in the model by adding the variable with the greatest improvement in R², i.e. SRT measured unaided in speech-shaped noise (R² change: 0.02). The variable age was included in the third step of the analysis, which significantly contributed to the model, with an R² change of 0.04. Additionally, the frailty index FI-CD₂₃, as a measure of the health status of the participants which might not be reflected by the numerical age, was included (R² change: 0.07). In a fifth step, the summarised mental-health score from the SF12 questionnaire was included, again contributing significantly to the model (R² change: 0.01). Finally, the group variable willingness to use HAs contributed significantly to the model (R² change: 0.063). Regression coefficients indicated that lower (better) PTA4 and SRT50 were related to lower HHIE/A scores. Higher age was associated with lower HHIE/A scores, while higher frailty and lower mental health scores were associated with higher HHIE/A scores. Participants who stated their willingness to use HAs were more likely to have higher HHIE/A scores compared to normal-hearing participants and to those unwilling to use HAs.

Table 3. Best predictive model for HHIE/A ratings including six statistically significant predictors.

	R	R^2	adj R^2	SE	R^2 change	F change	df1	df2	Significance of change
PTA4	0.54	0.29	0.29	12.72	0.29	85.96	1	208	<0.001
SRT50	0.56	0.31	0.31	12.57	0.02	5.87	1	207	0.016
Age	0.59	0.35	0.34	12.24	0.04	12.12	1	206	<0.001
FI-CD23	0.64	0.41	0.40	11.65	0.07	22.63	1	205	<0.001
Mental health	0.65	0.43	0.41	11.56	0.01	4.31	1	204	0.04
HA willingness/use	0.68	0.46	0.44	11.28	0.03	11.05	1	203	0.001

4. Discussion

The aim of the present study was to determine the characteristics of participants that influence self-assessed ratings of experienced hearing handicap in the HHIE/A questionnaire. Following a patient-centered approach it was hypothesised that besides the sensory deficit itself additional factors of the person and its living situation have to be taken into account, when providing audiological care and designing it around the individual’s needs. A stepwise regression analysis showed different personal factors that contribute to the experienced hearing handicap. In addition to the pure-tone hearing thresholds and speech recognition in noise, age, frailty, mental health and HA willingness/use were related to HHIE/A scores.

4.1. Audiological assessment

Pure-tone HL, characterised by PTA4, had the most explanatory power in the regression model (Table 3). PTA4 accounted for 29% of the variance in HHIE/A scores in this study, which is comparable to earlier findings (Weinstein and Ventry 1983; Newman et al. 1990). Nevertheless, this is not conclusive, as a lower predictive power of pure-tone HL (Moser, Luxenberger, and Freidl 2017b) and only small associations between HL and HHIE/A scores have also been reported before (Dalton et al. 2003; Eckert, Matthews, and Dubno 2017; Alhanbali et al. 2018). Some studies even found no significant association between HHIE/A scores and HL (Alpass et al. 2001; Preminger and Meeks 2010). A possible reason for this might be the group of participants tested. If all participants were hearing impaired or had narrow distributions of HL, the relation of HL (often scored as PTA4) and HHIE/A was lower or not significant at all. In a cohort with broad HL distributions, as described here, the influence of pure-tone thresholds in the perceived handicap is important but offers no complete variance explanation.

All variables in the model that were added after PTA4 had lower predictive power, which is partly due to the cross-correlation of the other predictors with pure-tone HL (e.g. speech recognition, age). Nevertheless, in addition to the PTA4, the unaided speech recognition in noise score was also a statistically significant predictor of HHIE/A scores in this sample, explaining an additional 2% of the variance in HHIE/A scores. Correlations of different speech recognition tests with HHIE/A outcomes have also been reported for speech recognition of words (Weinstein and Ventry 1983; Chang et al. 2016), of numbers (Alhanbali et al. 2018), and of sentences (Bertoli, Probst, and Jordan 1996). Furthermore, aided speech recognition in 4-talker babble noise predicted HHIE scores in hearing-impaired persons, while PTA did not significantly contribute to the regression model (Preminger and Meeks 2010). To verify whether aided speech recognition or different background noises influence the relation between speech recognition and HHIE/A score, the different types of speech-recognition scores and supra-threshold auditory perception assessed in this study were entered alternatively into the model in the second step of the stepwise regression. For the data analysed, unaided SRTs better predicted HHIE/A scores compared to aided SRTs with the same stimuli, or to any other SRT outcome. This also holds true if SRTs were selected according to the daily-life aiding situation of the participants, being aided SRTs for HA users only, and unaided SRTs for NH, and participants with HL but no HA use in daily life. This could be caused by the different aiding in the experiments, because participants did not wear their own HA during speech-recognition testing and were not accustomed to the HA provided. Another reason might be the questionnaire responses themselves. Participants were instructed to rate their regularly experienced conditions but, as some time might have passed between the telephone call and the self-assessment of the questionnaire, which was sent via mail, it remains an open question as to whether HA users rated the aided or the unaided condition. For speech recognition in noise (high- and low-context sentences), Eckert, Matthews, and Dubno (2017) found that speech recognition and subjective hearing handicap correlated more strongly if audibility was controlled in the analysis (by using the difference between importance-weighted prediction and speech recognition scores), instead of controlling for PTA (residualised scores). Whether this would also hold true for this study could not be resolved using the available data set.

4.2. Age, frailty and health-related QoL

In a third step, age was included in the model, and contributed significantly, explaining 4% of the unique shared variance with HHIE/A scores. HHIE/A scores did not follow a trend per 5-year age group (see Table 2 for the descriptives). Nevertheless, higher age was associated with lower HHIE/A scores when included after the HL in the regression model, contradicting the literature findings that indicate an increase in subjective hearing handicap with age (Dalton et al. 2003; Chang et al. 2016). Other studies showed no significant relation between HHIE/A and age of the participants (Eckert, Matthews, and Dubno 2017; Moser, Luxenberger, and Freidl 2017b). No literature evidence was found for a negative relation of age and HHIE/A scores, but for age and SSQ ratings, a similar trend was reported for participants aged 70 and older (v. Gablenz et al. 2018). In a cohort of elderly, first-time HA users tracked over 1 year after aiding, Malinoff and Weinstein (1989) suggested that the perceived hearing handicap does not increase in a uniform manner, but with some adaptation steps. However, as the participant sample described here consisted of persons with a wide range of HL and HA history, it is not likely that such adaptation steps to the individual HL and aiding are responsible for the observed effect. A more likely explanation for the negative relation between HHIE/A scores and age is a bias in the sample due to the recruiting process. During the measurement phase, the experimenters had the impression that younger participants volunteered to participate in the study because they experienced health-related and/or hearing deficits. The elderly age groups, conversely, seemed to be healthier and more active than their peers.

To examine the impression of selection bias, and to characterise the participants’ health status independently of age, the FI-CD was calculated retrospectively as a measure of frailty. The lack of age-dependency in the FI-CD data contradicts literature findings of frailty indices that indicated an increase of frailty with higher age (Rockwood et al. 2005; Searle et al. 2008; Kiely, Cupples, and Lipsitz 2009), but support the impression stated above that younger participants were more likely to have health deficits and elderly participants were healthier than their peers. Comparing the two approaches FI-CD₂₈ and FI-CD₂₃ (Figure 1), the uniform decrease of FI-CD₂₃ relative to FI-CD₂₈ points towards hearing difficulties being present over the whole age range. The FI-CD₂₃ was entered into the regression model in a fourth step and explained a further 7% of the variance in HHIE/A scores. Available evidence about the associations between perceived hearing handicap and frailty is rare. HL indicated by pure-tone audiometry was found to be associated with an increased risk of frailty, independently of age and other demographic factors (Kamil et al. 2016). Other health-related variables comparable to, or part of, a frailty assessment (e.g. multimorbidity, high blood pressure) did not predict HHIE outcomes (Eckert, Matthews, and Dubno 2017; Moser, Luxenberger, and Freidl 2017a). Frailty as applied by Kamil et al. (2016) was often restricted to physical abilities, whereas the FI-CD₂₃ applied also included subjective ratings, cognitive abilities, and dexterity, leading to a more global approach. However, it does not include vision impairments, and visual acuity was not identified as a single, additional significant variable in the regression model.

Alternatively to the FI-CD₂₃, the general-health rating adapted from the SF12 questionnaire (which is also part of the FI-CD₂₃) was entered into the regression model, but had less predictive power than the FI-CD₂₃ and did not contribute significantly to the model (data not shown here). This was rather unexpected, as the relation of HHIE/A scores to health-related QoL measured with the SF36 or SF12 questionnaire has been reported, indicating that higher HHIE scores are related to poorer physical health (Alpass et al. 2001). Nevertheless, mental-health scores adapted from the SF12 questionnaire were entered in the model, predicting additionally 1% of variance in HHIE/A scores. Higher mental-health scores were significantly associated with a lower handicap indicated by HHIE/A scores. In comparison to literature values, the proportion of the variance explained might be lower in the current analysis, because the regression model already accounted for HL and age, which are related to QoL ratings and therefore hinder to distinguish between effects (Nordvik et al. 2018). Associations between HHIE/A scores and mental-health ratings have also been described as being smaller compared to general health scores (Alpass et al. 2001). In some studies, HHIE/A scores were interpreted as hearing-related QoL (Maeda et al. 2016), which might not be precise enough, since not all aspects of health-related QoL are included in the HHIE/A items (Stika and Hays 2015).

4.3. Aiding status

The group variable distinguishing participants who stated their willingness to use HAs and those unwilling to use HAs or that had normal hearing was also a significant predictor, and explained 3% of variance in the HHIE/A scores. The coefficients indicated that persons who were willing to use HAs perceived a higher hearing handicap as indicated by higher HHIE/A scores. This implies that a considerable amount of perceived handicap might be required to accept the necessity of aiding; this corresponds to recent findings showing the experience of higher handicap of HA users in relation to non-users (Maeda et al. 2016). Furthermore, Gopinath et al. (2011) reported a regression model in which a HHIE score (screening version) of 8 and higher was a significant predictor for the incidence of HA ownership and use as dependent variables. Alternatively to the HA willingness/use, different HA-related variables (e.g. ownership, use, duration, satisfaction) were added in the sixth step, but mostly did not produce additional explanation of the variance and sometimes resulted in less predictive power in the model.

4.4. Limitations

There are some limitations to the data reported in the current study. Despite the potential bias in the recruiting mentioned above, the validity of the FI-CD needs discussion. Compared to the design of Searle et al. (2008), the procedure to calculate FI-CD was modified, due to the lower number of variables, and the retrospective calculation. Therefore, a cut-off point of deficits to differentiate between frail- and non-frail participants was not applied or defined here. Our approach might over- or under-rate the participants’ frailty status, as no standard measurement for general frailty was available. Nevertheless, the approach used to calculate cumulate deficits is a commonly used frailty measure, and was reported to have high validity and reliability (Dent, Kowal, and Hoogendijk 2016).

Furthermore, the provision of HAs in this study has a disadvantage. HAs were provided to all participants whose pure-tone HL was high enough to fulfil the German guidelines for HA provision irrespective of their hearing device status. Sixty-eight participants were already aided and therefore familiar with amplification. During part of the SRT measurements, all participants wore identical hearing aids provided and fitted to the individual HL by the experimenter. Moreover, all participants were aided during the non-audiological tests with their own HAs (if available) or the provided devices, having only a short time to adapt to the new listening situation (approximately 15 min) before the testing protocol continued. In this study, the unaided speech recognition scores were used in the regression model, which are therefore not influenced by HA provision. Nevertheless, due to the usage of HAs during the motor and cognitive testing, which was included in the calculation of the frailty index, a marginal influence of insufficient acclimatisation to the HAs cannot be precluded.

Another limitation to the outcomes of this study is that the HHIE/A scores might be underrated because of the self-assessment character of the questionnaire. Significant others were found to rate higher HHIE/A scores with regard to the related person than the self-assessed scores of the hearing-impaired persons (Chmiel and Jerger 1996). This external perception was not included here.

Moreover, the sample aged from 55 to 81 years required two different versions of the questionnaire (HHIE and HHIA). In the analysis, these were joined, as in other studies (Preminger and Meeks 2010; Stika and Hays 2015). This procedure might only have a minor impact on total scores, as only three questions differed between the two versions and no apparent changes were found in the regression model when the HHI total score, including the 22 common questions of the E/A versions, was used as dependent variable (data not shown here). Because no significant difference was found between the subscales, HHIE/A total scores were used as dependent variable in the analysis.

Although the HHIE/A instrument was regularly used to measure hearing-related QoL in many studies and in the rehabilitation process of persons with hearing impairment ( i.a. Aazh and Moore 2017; Chang et al. 2016; Öberg, 2016), the initial process of validation of the questionnaire lacks some information (Lazzarotto, Baumstarck, and Auquier 2016). The HHIE/A questionnaire was developed from expert opinion instead of patient perspective. In our study the missing data indicate that not all questions are relevant to all people. To determine the quality of targeting and dimensionality, two recent studies analysed the properties of the HHIE/A using modern statistic approaches (Cassarly et al. 2020; Heffernan, Weinstein, and Ferguson 2020). Based on the outcomes, both publications independently developed new uni-dimensional scales by removing items and refining the scoring of specific items of the HHIE/A. Both approaches were applied to the dataset presented here and the newly calculated HHIE/A sum scores were used as dependent variables in the regression model. This did not result in substantial changes to the model. Because there is no evidence regarding which potential effect the revision of HHIE/A scoring might have on the German version of the questionnaire, results were reported here for the original scoring method.

5. Conclusion

Overall, the stepwise regression analysis revealed that in addition to the pure-tone hearing thresholds and speech recognition in noise, other variables are related to HHIE/A scores, including age, frailty, mental health, and HA willingness/use. This result supports existing evidence that the handicap deriving from a HL is not determined by the HL alone, but dependent on other personal factors. This conclusion demands patient-centered care in audiology and thus considering additional health-related factors when evaluating the impact and handicap that a HL may have on a client’s life.

Acknowledgments

The authors thank Annäus Wiltfang, Olga Schwarz, and Annika Morgenstern for conducting the measurements, and Alina Baltus for the expert’s report on the neuropsychological test outcomes. Parts of the data were presented at the Fourth International Conference on Cognitive Hearing Science for Communication (Linköping, 2017). The procedure of retrospective frailty index calculation was presented at the annual meeting of the German Audiological Society (DGA; Halle (Saale), 2018). English language support was provided by http://www.stels-ol.de/.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Supported by the research professorship “Audiologie, Kognition und Sinnesleistungen im Alter (AKOSIA)”, in cooperation with the KOMUS study financed by Sonova AG (Stäfa, Switzerland).